More precisely delivering higher doses to the target volume at each treatment fraction, while limiting exposure to surrounding normal tissue, is the key goal for external beam prostate radiotherapy. Efforts to quantitatively assess this process have shown that a variety of uncertainties in the treatment setup including patient positioning, patient organ motion, characteristics of the treatment beam and operator variability issues make achieving this goal difficult. Previous efforts on this grant and by others have attempted to use the registration of planning day 3D computed tomographic images and intra- treatment megavolt portal images to try to account for some of these uncertainties, sometimes augmented by ultrasound imaging to try to estimate organ motion. While results are at times promising, these technologies are limited in their utility. Furthermore, while the ability to deliver more precisely- shaped plans has advanced significantly due to the advent of intensity modulated radiotherapy (IMRT) systems, the advanced development of image-guidance systems necessary to accurately deliver these more complex plans has been lacking. However, IMRT treatment systems with kilovolt (kV) 3D cone-beam CT (3DCBCT) imaging on the same platform have been developed, promising high quality imaging of both bone and soft tissue at each treatment fraction. We propose here to develop an advanced image analysis strategy that will permit the fully automatic nonrigid registration of planning day 3DCT images to 3DCBCT images acquired at each treatment day, while simultaneously and automatically segmenting the prostate, rectum and bladder in the treatment images. It is our view that this will ultimately facilitate the optimal, image-guided adaptation of an initial plan to each treatment day image, taking full advantage of these high resolution datasets. The approach will be validated using images formed from simulated deformations and evaluated using sets of patient images acquired at two different facilities over 6 weeks of treatment. [unreadable] [unreadable]